This will be the second study in the list from Krauss and Siri-Tarino's dairy-funded study on saturated fat (first below). This one is the Honolulu Heart trial. S-T & K used the 1984 paper but much more has been published on this dataset.
Honolulu Heart, part of a larger study called the NiHonSan study looking at Japanese people living in Japan, Hawaii and California determined what risk factors were present that were associated with heart disease. For reference, Framingham showed 12/1000 risk of CHD. 
NiHonSan showed significant differences among the groups, with Japanese in Japan having much lower risk even when compared with US residents with the same risk factor profiles (except that no Japanese had chol. > 260.
Honolulu Heart looked mainly at Nisei and Issei Japanese living in Honolulu and tried to find all of them. They did a single food frequency recall. Honolulu Heart showed dietary and serum cholesterol were big risk factors for CHD as was SFA.
There were some weird blips in the data though. As a percent of calories, the higher the SFA, the greater the heart disease, but those at the very lowest SFA consumption were also high in CHD. Cholesterol in the diet did not show this pattern.
The explanation that seems most obvious to me but without access to the baseline data I can't be sure, is that the ones at the lowest end of SFA consumption were likely drinkers whose 24 hour dietary recall might have included very little food and lots of liquid.
Alcohol was a significant risk in the study. The prevalence of alcoholism in Japanese men on Oahu around that time was between 4 and 10 percent, depending on what source you read. Likely all estimates of alcoholism are slightly underreported.
Important methodologically though, is the fact that the analysis of S-T & K only computed the delta between members of individual studies and then averaged them. They did not run a comparison between the rates of CHD in Honolulu compared to Western Electric. for example.
Honolulu Heart later published data showing increased rates of overall mortality in lower consumers of SFA and lower rates of stroke in higher SFA consumers. The associations largely disappeared when calories from alcohol were accounted for however.
Honolulu Heart looked at a very homogenous population with little variability. The fact that it found significant differences is surprising. The big picture of NiHonSan shows the overall pattern of decreased SFA is beneficial for CHD. The anomalies are likely due to alchohol use.
Next time we will look at Boston Diet-Heart study.
